Maximum entropy inference method applied to shallow-water ocean waveguides.

Abstract

The principle of maximum entropy (MaxEnt) is used to analyze the uncertainty that is associated with determining geo-acoustic waveguide parameters for shallow-water ocean environments. Marginal probability distributions and statistics, such as the mean and standard deviation, for each geo-acoustic parameter are obtained using MaxEnt methods. Both theoretical and numerical components of the MaxEnt method are addressed. The method is applied to experimental acoustic measurements taken on the New Jersey continental shelf on a sand ridge where a priori information includes sound velocity measurements in the water column and chirp seismic reflection grams that provide geophysical ground truth on the seabed layering structure. The acoustic data used in the geo-acoustic inference result from a towed multi-frequency source. The effect of discretization choices for the geo-acoustic representation on the marginal probability distributions is investigated. The effect of changing a priori information (such as the assumed sound velocity profile in the water column) on the marginal probability distributions of the geo-acoustic parameters is also explored.